feat: consolidate TraceDecay V2 PR8-PR13 delivery - #421
Draft
ScriptedAlchemy wants to merge 4698 commits into
Draft
feat: consolidate TraceDecay V2 PR8-PR13 delivery#421ScriptedAlchemy wants to merge 4698 commits into
ScriptedAlchemy wants to merge 4698 commits into
Conversation
|
ScriptedAlchemy
force-pushed
the
codex/tracedecay-total-redesign-plan
branch
from
July 10, 2026 05:54
e50f8f5 to
4f5ab66
Compare
Owner
Author
|
@codex review |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 986f25cca6
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
ScriptedAlchemy
force-pushed
the
codex/tracedecay-total-redesign-plan
branch
4 times, most recently
from
July 11, 2026 16:52
5862bd6 to
53502de
Compare
ScriptedAlchemy
force-pushed
the
codex/tracedecay-total-redesign-plan
branch
4 times, most recently
from
July 23, 2026 17:06
642a33a to
5f2f610
Compare
ScriptedAlchemy
changed the title
ScriptedAlchemy
force-pushed
the
codex/tracedecay-total-redesign-plan
branch
from
July 24, 2026 17:16
76dee0f to
0258866
Compare
ScriptedAlchemy
added a commit
that referenced
this pull request
Jul 27, 2026
Record the 2026-07-27 audit in the authoritative V2 status files. Three delivered claims (Plan 16 worktree/store identity, Plan 38 cursor repair, Plans 18/23 end-to-end sanitization) lost their tests in 9e3ca9f. Filed as verification-coverage corrections, not retractions: the assertion is missing, the behavior is not withdrawn. Also record the six instances of gates attesting to something they never checked, the libtest empty-filter asymmetry that hides them, the six product defects fixed tonight with their SHAs, and Plan 11's corrected open-gap list including two items the widened accessibility matrix surfaced. None of tonight's commits has CI validation: PR #421 has been conflicting since 05:13 UTC, so no CI has run since 01:24 UTC. Every fix is recorded as locally verified only.
ScriptedAlchemy
force-pushed
the
codex/tracedecay-total-redesign-plan
branch
from
August 1, 2026 18:32
062285a to
b73f8bf
Compare
…-total-redesign-plan
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The first query against a restored generation paid the whole cold path: the sealed decode (which re-mints every file's exact-extraction authority with a canonical SHA-256 over every chunk, then repeats the full canonical validation sweep), plus the exact-admission sweep, record lookup indices, lane owners, and test-attribution join — all deferred behind OnceLocks and charged to whichever request happened to arrive first. That is O(store) on the serving path, violating the invariant in docs/SERVING-PATH-PERFORMANCE. Activation now owns that work: - mount runs the worktree open on the blocking pool instead of an async runtime worker, and warms every per-generation derivation there; - reconcile completion warms the generation it just published; - the publication store gains a single-flight decode barrier. The cache lock is never held across a decode, concurrent callers for the same generation share one decode instead of duplicating it, and a publication landing mid-decode retires the superseded result via an epoch guard; - the active generation is pinned outside the LRU, so cursor traffic over superseded generations can never evict it. Fail-closed is unchanged: only success is memoized, so a corrupt store still runs the complete check and errors on every request, and warming is best-effort — it is a pre-warm, never a gate. Measured on a release build over a 1.4K-file fixture: first-query cost 3.57s -> 88ms. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…-total-redesign-plan
Cold generation decode spent 40%+ of its cost in ExactExtractionAuthorityV1 mint/restore: every chunk was serialized to canonical JSON and SHA-256'd one at a time on one thread. Fan the per-chunk sweeps (mint, validate_all, admit_all, and the per-chunk leg of CodeFileChunksV1::validate) across the pool above a 16-chunk threshold. Digests are independent, results are collected in chunk order, and the reported failure is still the lowest-index one, so callers observe exactly what the sequential short-circuit produced. Cut the allocation churn inside canonical writing: strings and object keys now stream their JSON escapes straight into the sink instead of allocating a String per value via serde_json::to_string, integers format into a stack buffer, and already-sorted objects skip the collect-and-sort Vec. Output stays byte-identical, covered by escape-class and number equivalence tests against the rendering they replace. Measured on a 1201-chunk fixture (release, min of 25): mint 10.98ms -> 2.0ms; the canonical change alone takes the single-threaded sweep 10.98ms -> 9.6ms. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…-total-redesign-plan
`tracedecay_runtime_core::ports::registered_schema` fails closed until the composition root installs the real schema, so every test process that opened a profile- or session-scoped shard without registering died with "no registered global/session schema installer is registered". That was the largest CI cluster: ~240 lib tests across six crates. Register the real installer at each crate's *fixture funnel* rather than at every test, following the precedent already set by `tests/common/mod.rs` and `src/dashboard.rs`: - `tracedecay-global-db` self-registers inside `open_registered_test_database`, `RegisteredGlobalDbHarness::open`, both harness `open` constructors, and the LCM schema read-only fixture. That covers every downstream crate that builds fixtures on `HostAdmissionTestRuntimeV1` / `RegisteredGlobalDbTestRuntime`. - `tracedecay-migrate`, `tracedecay-usecases`, `tracedecay-dashboard-api`, and the root crate register at the handful of fixtures that reach `Database::publish_test_runtime` directly. - `tracedecay-runtime-core` sits below the crate that owns the schema and can never install it, so its *own* `cfg(test)` build treats an unregistered port as an empty profile sidecar. Kernel fixtures only ever exercise the graph shard beside it. No other build reaches that arm. Two adjacent fixture defects surfaced once the port stopped short-circuiting, both in `open_registered_test_database`: schema DDL ran on `Database::conn()` (the retained *reader*, so the migration SQL channel answered `WriterUnavailable`), and the mode was hard-coded to `Initialize` even for a store an earlier stage of the same test had already materialised. Both fixed. Production stays fail-closed, and `crates/tracedecay-global-db/tests/registered_schema_fail_closed.rs` now proves it from a process that links the kernel as an ordinary dependency and registers nothing. Lib tests, `--all-features`: 305 -> 157 failing (148 fixed, 0 newly failing); the schema-installer error is gone from every measured crate. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
`HostBundleError::StorageFailure` had ~115 construction sites across the host bundle lifecycle that all rendered the identical sentence, so a failed install or uninstall reported nothing an operator or a bug report could act on. Carry the raising site in the variant. `host_bundle_storage_failure!()` stamps `file:line` at the call site, keeping the payload a `&'static str` so the enum stays `Copy`/`Eq`. Tests that assert the whole error value now compare against a named sentinel instead of a second macro expansion, which would carry the assertion's own line. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
`tracedecay uninstall --agent codex` failed with a bare storage failure and rolled the whole transaction back, so uninstall was a no-op that reported an error. Codex's `host_component_registration` gated on the deployed plugin *source* bundle: with `.codex-plugin/plugin.json` on disk it never returned `Missing`, only `Current` or `Repairable`. But that bundle is an artifact-layer deployment which `deactivate_deployed_host_registration` deliberately leaves behind, so after a correct deactivation the component still read `Repairable` while the uninstall verify demands `Missing`. Gate on the registration surface instead, matching Claude's residue contract: the `[hooks.state]` trust keys, `[mcp_servers.tracedecay]`, the `[plugins."tracedecay@..."]` activation records, and the personal marketplace entry — exactly what deactivation clears. A staged but never activated bundle still reads `Repairable`, since its marketplace entry survives. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…-total-redesign-plan
The three projection failures blamed on the migration removal trace to the earlier recall fix that narrowed derived logical copies to genuine re-emissions: a parent-message link is threading, not a copy. The tests encoded the old every-reply-is-a-copy semantics and are re-contracted; copy persistence, restart idempotency, and back-off coverage are unchanged in strength. The retrieval candidate fixture had aborted at seed time since its FK on session_turns went unsatisfied, so its expectations had never run; it now seeds parent turns, scopes the snapshot to the fixture session, asserts all in-window occurrences, and narrows three query-plan assertions that were unsatisfiable by query shape. One assertion stays red on purpose: the root-wide derived candidate query full-scans session_derived_evidence because its covering index leads with session_id - a real performance question, not a test defect. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
# Conflicts: # crates/tracedecay-migrate/src/manifest/runtime.rs # tests/storage_suite/migration_manifest_test.rs
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
This draft PR is the consolidated delivery branch for TraceDecay V2 PR8 through PR13.
The branch also contains the supporting rusqlite runtime cutover, worktree-aware incremental indexing, storage retention/Doctor integration, and direct product tests required by the canonical V2 plans.
Verification
cargo check --all-featurescargo test --all-featurescargo clippy --workspace --all-targets --all-features -- -D warningscargo fmt --all -- --checkCurrent status: implementation is still being consolidated from the shared checkout and CI is red. This PR remains draft until the working tree is committed, broad verification is stable, and the historical conventional-commit blocker is resolved. It must not be merged in its current state.
Safety boundaries
GitHub review ingestion is read-only. Semantic indexing never blocks exact/lexical/graph retrieval and publishes only complete compatible generations. Local coordination logs, private transcript evidence, secrets, and ad hoc artifacts are excluded from commits.